Jev in production › Agent context and memory

jev-skills

Vendor claims Jev decides which installed skills a Claude Code or Codex model sees each turn, cutting a measured 1,350-token-per-turn baseline to zero (author).

Open on GitHub ↗

1,350 tokens per turnmeasured against a baseline, as published by the source
Use
Agent context and memory
Industry
Developer tools
Form
Open-source tool
Stage
Announced
Plugs into
Claude Code
Listed
2026-09-25
Found via
github
Repository
eran-broder/jev-skills
Stars
6
Forks
1
Last push
2026-09-23
Language
TypeScript
License
MIT
jev-skills screenshot
media/poster.png in the eran-broder/jev-skills README, MIT; shown from GitHub.

The README opens with

A plugin for Claude Code and Codex that takes every skill out of the context window and lets a calibrated classifier, TypeSafe's Jev, decide on each turn which ones the model should see.

Always-on context cost: 0 tokens. Library size limit: none that the model can feel.

72 seconds, every transcript and number from a real run. Source for the video is in media/video.

Every skill you install costs you on every turn, whether you use it or not. Claude Code lists each skill's name and description in the model's context. Codex does the same and stops at about two percent of the context window, then silently drops the rest.

Badge

For the project's own README, linking back here:

Listed in Jev in production

Also used for agent context and memory